SPSS Statistics

 View Only
Expand all | Collapse all

SPSS(v29.0.1.1) issues with TURF analyses

  • 1.  SPSS(v29.0.1.1) issues with TURF analyses

    Posted Wed November 20, 2024 09:30 AM

    Hi all!  When I run a TURF analysis, it seems like it's doing some sort of stepwise analyses where it adds and then drops flavors to get to "best reach".  In addition, if I re-run the analyses, I often don't get the same results (whether I use only syntax, only the drop down menu, or both). Is anyone else running into this problem and/or has a solve for this?  I am running on v29.0.1.1 (subscription). 

    A million thanks! Alicia.



    ------------------------------
    Alicia Elsholz
    ------------------------------


  • 2.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Wed November 20, 2024 09:50 AM
    TURF proceeds by adding the best variables (choices) one by one, but it also removes previously added variables when the addition of a new one makes one previously chosen no longer optimal. 

    You could think of this as similar to stepwise regression when using the full stepwise method: a variable can be removed due to an earlier added variable not being optimal in the presence of a newly added variable.  Naive TURF implementations that I have seen neglect the removal step, which results in less than optimal results. 

    As for varying results with TURF, when there are a lot of input variables, the runtime can become very large (exponentially increasing).  So if there are a lot of choices, TURF uses a tournament method (for which I have a patent - I am the author of TURF).  It runs heats on randomly constituted variable subsets and takes the top few choices from each heat and runs a final round over these choices.  This could cause the final winners to vary a bit, but this would normally happen only when the results are so similar that the exact choice has negligible effect on the quality of the outcome.

    Heats are only used for large problems, but you can disable this feature on the Heats subdialog.  That should give you repeatable results except in the very rare case where multiple choices are exactly tied.

    --





  • 3.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Fri November 22, 2024 07:59 PM

    Thank you so much for your response, Jon!  I'm glad to know it wasn't something I was or wasn't doing and that it was operating as it should.  Additionally, I turned off the "heats" capability, and it did, in fact, eliminate the variability in the results -- what a relief!  One additional thing I would love your help with, is in visualizing results.  For example, using dummy data/results below, I would like to add flavor names to the graph below but I'm not sure how to do that when flavors are removed (see below for graph and table).

    For example, I would label "1" on the graph as "Strawberry Lemonade" based on the results in the table, and "2" as "Peach Lemonade"...but when I get to "7" a bunch are added and removed -- even "1" Strawberry Lemonade is removed.  So I'm not sure how best to visualize and report results if I want to display 7 flavors but realistically only recommend 1-3 for development -- usually the first 3 items.

    Any guidance you can provide here is greatly appreciated, Jon!  Thanks!!



    ------------------------------
    Alicia Elsholz
    ------------------------------



  • 4.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Fri November 22, 2024 09:59 PM
    Nice chart :-)
    Glad you are making progress on this.
    The standard chart is designed to show the incremental improvement as you add products, so you wouldn't want to mess that up.

    The best way to add more detail would be to edit the x-axis labels to show the products included at each size listing them one per line.  You would need to enlarge the chart, shrink the axis font, and abbreviate the names/  You would only show the products at each size, dropping the added/removed/retained information.  Here's a mockup of what I am thinking of.  I'm not sure that this will come through on the site, but I can send you the image by email if it doesn't (send me a message to jkpeck@gmail.com)

    image.png

    Using the Chart Editor, you can make all these edits. Shift-enter gives you the line breaks.


    --





  • 5.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Mon November 25, 2024 10:08 AM

    Thanks so much, Jon!  I understand what you are saying -- the label of each data point would include only the items that are kept so, with respect to my dummy data, even though "1" = strawberry lemonade, and "2" = strawberry lemonade + peach lemonade, when I get to "7" I would exclude strawberry lemonade because it was removed.  Makes total sense. 

    One last clarifying question.  If a client is only interested in adding 1-2 flavors, is it still acceptable to recommend strawberry lemonade and peach lemonade since those two flavors have the greatest reach -- even though strawberry lemonade is removed at "7"?

    Thanks! Alicia.



    ------------------------------
    Alicia Elsholz
    ------------------------------



  • 6.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Mon November 25, 2024 12:06 PM

    Hi Alicia,

     

    Remember what the intent of your exercise was, and what you are telling the model to do. If your goal was to pick the top three or four optimal flavors to maximize reach, then you have clear direction: Strawberry Margarita, Strawberry Lemonade, Peach Lemonade, and Watermelon. The incremental contribution is steady (about +10 ppts) until you get to group 6. And by the time you get to group 7, you're basically telling the model to ignore all previous combinations, and to go out and find new combinations that could do even better, which is why you're getting a rather odd result.

     

    The way you are displaying the results is great, but I'd just go with the reach component as frequency is somewhat harder to communicate (unless you included a volumetric measure, it is just a share of votes). If you had included a volumetric component, e.g., consumption frequency, you could add that in as a weighting variable or tie breaker. I'm guessing that these flavors all probably pretty well. And I'm suddenly getting thirsty!

     

    Cheers,

     

    Bob Walker

     

     


    A picture containing stop  Description automatically generated

    Robert W. Walker

    CEO & Founder

    23 S. Main St.·  Norwalk, CT 06854

    Tel. +1.203.255.0505

    https://usemotion.com/meet/rwwalker/meeting

     

     

     






  • 7.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Mon November 25, 2024 12:17 PM
    I wouldn't characterize the result as odd.  It's just a consequence of overlap of other flavors.  I have seen cases with much more dropping out of variables that started out strong.

    Note also that the reach difference between the best 7-flavor bundle and the one with the no-dropped flavor may be very small.


    --





  • 8.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Fri December 06, 2024 10:07 AM

    Thank you so much, Bob!  Super helpful and love the reco to focus on reach. Hope you have a lovely weekend!



    ------------------------------
    Alicia Elsholz
    ------------------------------



  • 9.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Mon November 25, 2024 12:16 PM
    Yes.  You have to edit the labels manually, unfortunately.  I have considered stacking the choices in the way I described as a TURF option, but it would be very difficult to get the alignment, font, etc right in the general case.

    As for your 1-2 flavors question,  what happens with larger bundles is irrelevant.  Dropping a flavor at step 7 is just a consequence of having chosen the first six, that particular 7th choice is not as good as something else.  So there is no reason to let it affect earlier best bundles.


    --





  • 10.  RE: SPSS(v29.0.1.1) issues with TURF analyses

    Posted Fri December 06, 2024 10:11 AM

    Thank you so much, Jon!  I feel so much more confident about the results and my recos now!  A million thanks! Alicia.



    ------------------------------
    Alicia Elsholz
    ------------------------------